Online Appendix to Quality Disclosure Programs and Internal Organizational Practices: Evidence from Airline Flight Delays

Transcription

1 Online Appendix to Quality Disclosure Programs and Internal Organizational Practices: Evidence from Airline Flight Delays By SILKE J. FORBES, MARA LEDERMAN AND TREVOR TOMBE Appendix A: Identifying Reporting Technology at the Aircraft Level We only have information on the reporting technology that each airline uses beginning in March We can safely assume that carriers that are manual reporters in 1998 are also manual reporters in prior years because, if their planes do not have ACARS in 1998, they would not have ACARS prior to For the carriers that use a combination of manual and automatic reporting, we do not know which or how many of their aircraft are equipped with ACARS. Therefore, we develop an approach for identifying specifically which of their aircraft is likely to be reporting manually. We use the same method to verify that carriers that are automatic reporters in 1998 also reported automatically from Specifically, we identify a manual plane based on whether delays on the flights carried out by this plane are rounded in a way that is similar to how the manual reporters appear to round their delays. We are able to do this because we can track planes in our data by tail number. Thus, for each aircraft in each year of our data, we calculate the fraction of its flights in that year that have a reported arrival delay of exactly zero. We then compare the distribution of this plane-year level variable across airlines which report their on-time data in different ways. Table A1 below shows the distribution of this variable for the 10 airlines who reported in The 95 th percentile of the distribution of this variable for American Airlines (an automatic reporter) is which indicates that only about 1 percent of American s planes arrived with a delay of zero minutes more than 4.6% of the time. Contrast this to America West (a manual reporter): 50% of its planes landed with a reported delay of zero more than 4.7% of the time. Southwest is an outlier with the 50 th percentile of its distribution being 11.7%, far higher than any other airline s. Based on this table, we categorize any plane that has reported delays of zero for more than 5% of its flights in any year as a manual plane for every year of our sample. We see this as a conservative approach to identifying manual planes since it classifies a plane as manual based on it meeting the criteria described above in only a single year. Using this approach, we classify 85 of Continental s 441 planes (19%) and 22 of TWA s 241 planes (9%) as manual. Table A1 Likelihood of a Plane Landing with Exactly Zero Delay (By Carrier, in 1998) 50 th 75 th 90 th 95 th 99 th Reporting Status in 1998 Alaska Manual America West Manual American Auto Continental Combo Delta Combo Northwest Auto Southwest Manual TWA Combo United Auto US Airways Auto

2 Appendix B: Two-Minute Early Results Table B1 Probability of Arriving Two Minutes Earlier than Predicted, , All Reporting Carriers [10,11) min 0.006*** [11,12) min 0.008*** [12,13) min 0.011*** [13,14) min 0.007*** [14,15) min 0.010*** [15,16) min 0.013*** [16,17) min 0.048*** [17,18) min 0.012*** [18,19) min 0.005** [19,20) min 0.012*** [20,21) min 0.014*** [21,22) min 0.010*** [22,23) min 0.016*** [23,24) min 0.005** [24,25) min 0.016*** 25 min 0.015*** P-Values from Hypothesis Tests of Equality of Coefficients (two-tailed) Bin16=Bin Bin16=Bin Bin16=Bin Prob(2 min early) in group 0.08 Notes: Standard errors are in parentheses and clustered at the level of the arrival airport-day. Coefficients represent the change in the probability of a flight arriving two minutes earlier than predicted relative to flights with predicted delay less than 10 minutes. The regression contains 3,326,681 observations. ** significant at 1%. * significant at 5%. + significant at 10%.

3 Table B2 Probability of Arriving Two Minutes Earlier than Predicted, Carrier-Specific Estimates Carrier Coefficient on [16,17) Bin Prob(2 min early) # (%) Flights in [16,17) Bin Reporting Technology Bonus Program During American 0.018** ,013 (0.80%) Automatic N (0.006) Alaska 0.049*** ,195 (1.2%) Manual N (0.010) Continental 0.120*** ,213 (0.84%) Combination Y (0.009) Delta 0.026*** ,312 (1.1%) Combination N (0.004) America West 0.058*** ,492 (1.1%) Manual N (0.009) Northwest 0.024*** ,675 (0.79%) Automatic N (0.006) TWA 0.119*** ,485 (0.84%) Combination Y (0.011) United 0.013** ,421 (0.91%) Automatic N (0.005) US Airways 0.012** ,437 (0.95%) Automatic N (0.005) Southwest 0.137*** ,603 (0.68%) Manual N (0.006) Notes: The second column of the table shows the coefficient on the [16,17) bin for each airline. The coefficient represents the change in the probability of arriving one minute earlier than predicted for flights with predicted delay between 16 and 17 minutes, relative to flights with predicted delay less than 10 minutes. The third column shows the fraction of each airline s flights that arrive two minutes earlier than predicted. The fourth column shows the # (%) of each airline s flights that are predicted to arrive between 16 and 17 minutes late. The regression contains 3,326,681 observations. ** significant at 1%. * significant at 5%. + significant at 10%.

4 Table B3 Probability of Arriving Exactly Two Minutes Earlier than Predicted, Carriers that Switch Reporting Technology America West Alaska Southwest Manual Period Automatic Period [12,13) min *** (0.008) (0.006) (0.004) (0.004) [13,14) min *** (0.008) (0.006) (0.004) [14,15) min 0.039*** (0.009) (0.007) (0.003) (0.004) [15,16) min * (0.009) (0.006) (0.003) (0.004) [16,17) min 0.022* *** (0.010) (0.007) (0.005) (0.004) [17,18) min 0.036*** *** (0.010) (0.007) (0.004) (0.004) [18,19) min *** 0.014** (0.010) (0.007) (0.005).. 25 min 0.022*** 0.005** 0.005*** 0.006*** (0.003) Notes: Standard errors are in parentheses and clustered at the level of the arrival airport-day. Columns display select coefficients from a single regression that includes carrier-arrival airport-day fixed effects and arrival hour and hub controls. Coefficients represent the change in the probability of a flight arriving two minutes earlier than predicted, relative to flights with predicted delay less than 10 minutes. The regression contains 2,904,668 observations. ** significant at 1%. * significant at 5%. + significant at 10%.

5 Table B4 Probability of Arriving Two Minutes Earlier than Predicted, Combination Reporters, Variation in Bonus Programs Continental TWA pre-bonus TWA post-bonus Delta Manual Automatic Manual Automatic Manual Automatic Manual Automatic [12,13) min 0.036*** 0.024*** *** 0.022*** 0.005** (0.006) (0.003) (0.020) (0.006) (0.013) (0.004) (0.003) [13,14) min 0.028*** 0.028*** *** 0.018*** 0.005** (0.006) (0.003) (0.022) (0.007) (0.014) (0.004) (0.003) [14,15) min 0.036*** 0.042*** * 0.037*** 0.020*** 0.010*** (0.007) (0.004) (0.021) (0.008) (0.016) (0.005) (0.003) [15,16) min 0.058*** 0.059*** *** 0.064*** 0.018*** 0.013*** (0.007) (0.004) (0.027) (0.007) (0.018) (0.006) (0.003) [16,17) min 0.156*** 0.104*** ** 0.229*** 0.119*** 0.045*** 0.018*** (0.010) (0.005) (0.029) (0.008) (0.023) (0.006) (0.003) [17,18) min 0.022** 0.023*** *** 0.010*** (0.007) (0.004) (0.027) (0.008) (0.016) (0.005) (0.003) [18,19) min *** *** 0.009*** (0.007) (0.004) (0.027) (0.008) (0.017) (0.005) (0.003) 25 min 0.030*** 0.022*** 0.027** 0.019*** 0.025*** 0.023*** 0.029*** 0.021*** (0.009) (0.003) (0.005) P-Values from Hypothesis Tests of Equality of Coefficients (two tailed) Bin16=Bin Bin16=Bin Bin16=Bin Prob(2 min early) Notes: Standard errors are in parentheses and clustered at the level of the arrival airport-day. Columns display select coefficients from a single regression that includes carrier-arrival airport-day fixed effects and arrival hour and hub controls. Coefficients represent the change in the probability of a flight arriving two minutes earlier than predicted relative to flights with predicted delay of less than 10 minutes. The regression includes flights by Delta, Continental and TWA on every other day between 1995 and 2000 for a total of 4,485,758 observations. ** significant at 1%. * significant at 5%. + significant at 10%.

6 Table B5 Probability of Arriving Exactly Two Minutes Earlier than Predicted Introduction of Later Bonus Programs American Airlines US Airways United Airlines Pre-Bonus Post-Bonus Pre-Bonus Post-Bonus Pre-Bonus Post Bonus [12,13) min 0.010** 0.012*** 0.011* * (0.004) (0.004) (0.005) (0.005) [13,14) min 0.017*** 0.010*** ** (0.004) (0.003) (0.004) (0.005) (0.005) [14,15) min 0.014*** 0.019*** 0.017*** (0.004) (0.003) (0.005) (0.003) (0.005) (0.005) [15,16) min 0.022*** 0.021*** 0.011* (0.004) (0.003) (0.005) (0.003) (0.005) (0.005) [16,17) min 0.030*** 0.029*** (0.004) (0.003) (0.005) (0.003) (0.005) (0.006) [17,18) min 0.019*** 0.009** 0.015** 0.007* 0.015* (0.005) (0.003) (0.005) (0.003) (0.006) (0.006) [18,19) min 0.010* 0.009** ** (0.005) (0.003) (0.005) (0.003) (0.006) (0.006) 25 min 0.011*** 0.010*** 0.016*** 0.013*** 0.010*** 0.013*** N (Years included in sample) 2,777,448 ( ) 2,101,260 ( ) 1,087,605 ( ) Prob(2 min early) P-Values from Hypothesis Tests of Equality of Coefficients (two tailed tests) Bin16=Bin Bin16=Bin Bin16=Bin Notes: Standard errors are in parentheses and clustered at the level of the arrival airport-day. Separate regressions are run for each of the carriers in the table. The regressions include arrival airport-day fixed effects and arrival hour and hub controls. Coefficients represent the change in the probability of a flight arriving two minutes earlier than predicted, relative to flights with predicted delay less than 10 minutes. ** significant at 1%. * significant at 5%. + significant at 10%. 6